Sydney eScholarship
Not a member yet
31878 research outputs found
Sort by
Advanced Learning-Based Approaches for 3D Realistic Perception
No full text3D realistic perception refers to the faithful ego-centric understanding of real-word object attributes and their precise spatial localization, while maintaining geometric and semantic consistency. In contrast to purely semantic 2D perception, it provides spatially grounded representations—such as bird’s-eye-view (BEV) maps, 3D curves, occupancy, and object trajectories—that are actionable for planning and control. Such perception can be derived from a variety of sensing modalities, including LiDAR, cameras, or multi-modal fusion approaches. Despite these advances, the field is impeded with inherent challenges such as occlusion, long-range sparsity, adverse weather conditions, and domain shifts. In safety-critical applications like autonomous driving and embodied intelligence, the success of 3D realistic perception ultimately depends on achieving high metric accuracy, robustness, and computational efficiency.
This thesis advances learning-based 3D perception for lane-line and dynamic-object understanding in challenging driving environments. To address the scarcity and annotation burden of LiDAR data, it introduces LiSV-3DLane, the first large-scale surround-view 3D lane dataset with enriched semantic annotations. Based on this resource, LiLaDet projects LiDAR geometry into a BEV representation for precise 3D lane recovery and broader applicability of LiDAR-based perception. To reduce the high cost of point-cloud processing, LaneCMKT transfers 3D cues from a LiDAR teacher to a monocular image student via cross-modal distillation, improving detection robustness under adverse conditions. Finally, BeXT (Bringing eXpertises Together) integrates complementary Visual Foundation Models (VFM) into a lightweight monocular encoder through expertise adapter pretraining and dynamic feature routing. Collectively, these contributions establish a pathway from LiDAR to scalable monocular deployment, enabling robust, efficient, and metrically faithful 3D realistic perception
Heart rate variability after spinal cord injury
No full textIn addition to motor and sensory dysfunction, a spinal cord injury (SCI) can lead to disruption of the autonomic nervous system (ANS). The ANS contributes to the regulation of multiple internal body functions, such as respiration, blood pressure, heart rate, and peristalsis. Disruption of such a system can lead to sometimes serious and often debilitating complications, including autonomic dysreflexia, orthostatic hypotension, bladder incontinence, cognitive impairment, pain, and fatigue. These complications can have a substantial impact on quality of life, affecting areas such as physical independence, relationships, work, and leisure.
To support improvements in autonomic function, advances in the assessment and treatment of autonomic dysfunction are needed. Among the available autonomic assessment measures, heart rate variability (HRV) is a cardiac autonomic biomarker that assesses the activity of the ANS
at the sino-atrial node of the heart. Its non-invasive and accessible nature, as well as its ability to assess vagal (parasympathetic) regulation of the heart has made HRV a very popular autonomic biomarker.
Despite this, the assessment of HRV is far from standardised, leading to highly heterogeneous results and difficulty implementing it into a clinical setting. Subsequently, the effect of interventions on HRV, namely non-invasive, non-pharmaceutical (NINP) interventions, after SCI is unclear. HRV biofeedback (HRV-F), in particular, is an under-researched NINP intervention in adults with SCI, despite demonstrating positive cardiovascular and psychosocial outcomes in other chronic disease populations.
In light of this, this thesis aimed to address some of these barriers associated with the use of HRV in adults with SCI. Key issues related to the assessment of HRV and the effect of NINP interventions on HRV, particularly HRV-F, are explored over seven chapters, including five manuscripts
Persistence Under Pressure: Understanding Connectivity, Detection and Conservation of the Eastern Pygmy Possum in the Urban Matrix
No full textUrbanisation continues to fragment native habitats and place pressure on species that rely on continuous vegetation for movement and survival. This thesis investigates how urban-edge fragmentation affects the eastern pygmy possum (Cercartetus nanus), a small, cryptic and urban-sensitive arboreal marsupial in northern Sydney. Using SNP-based genetic analysis, dynamic occupancy modelling, detectability assessments and a synthesis of long-term wildlife-crossing data, the research provides an integrated evaluation of the factors shaping persistence and connectivity in a highly modified landscape.
Nest-box surveys showed that the species persists within fragmented bushland remnants, and that detection improves when surveys incorporate key feed trees. Occupancy was strongly influenced by native vegetation extent and fire history, emphasising the importance of habitat connectivity and management for long-term persistence. Genetic analyses indicated subtle but emerging structuring among habitat patches, yet limited differentiation across a major arterial road, suggesting roads may not yet form absolute barriers. A review of road-ecology literature and 25 years of monitoring revealed little population-level assessment for crossing structure performance and very few records of this species using existing crossings. Landscape-resistance and least-cost-path analyses identified priority corridors, identifying the location for species-specific crossings in the study area.
Together, these findings highlight the need for mitigation that integrates analysis of genetics, occupancy and connectivity, to provide a replicable framework for managing fragmentation in urban landscapes
Managing Industrial B2B Platforms
No full textThis thesis comprises four chapters. Chapter One provides a “meta-theoretical” agenda by revealing
different paradigms or ways of thinking about platform ecosystems. Through the theoretical synthesis
of existing research across management fields, Chapter One clarifies the onto-epistemological
foundations of each paradigm and helps explain the varying expectations researchers place on the
platform ecosystem construct. Chapter Two offers a conceptual framework that theorizes the nature
of B2B platforms as meta-organizations. Building on a systematic review of existing literature, it
uncovers five dominant B2B platform archetypes (matchmaker, application marketplace, solution
enabler, consortium, and decentralized autonomous platforms). Further, it theorizes how the specific
design features of B2B platforms shape their governance models.
Chapter Three and Four empirically explore the dynamics of establishing B2B platforms. Drawing on
a qualitative, single case study of a B2B solution enabler platform, Chapter Three shifts the focus
from static platform attributes to the unfolding communication and coordination efforts that underpin
the management of B2B platform establishment. Drawing on the framing concept, this chapter
foregrounds the role of narratives and their strategic and temporal complexity in establishing a B2B
platform, offering insights into the socio-cognitive nature of B2B platform establishment.
Chapter Four builds on the same single case study, discusses the paradox of openness, the “black
box,” that most platforms face, and shows how firms operationalize strategic openness throughout a
platform’s lifecycle. It provides a process model of platform openness, shaped by highlighting how
platform identity acts as both a driver and outcome of openness decisions across technical,
economic, and socio-cognitive dimensions, indicating that B2B platform establishment is a pathdependent,
multi-dimensional, and configurational process
An Evaluation of ASIC’s Evolving Enforcement Priorities and Praxis
No full textEnforcement action is a key responsibility of the Australian Securities and Investments Commission (ASIC), Australia’s national financial regulator. ASIC has been criticised for its poor enforcement record against serious widespread or harmful misconduct impacting consumers and investors. Many public reviews of ASIC’s enforcement activities have been undertaken triggering recommendations for improvement. This thesis critically evaluates ASIC’s response to the recommendations by the Royal Commission into Misconduct in the Banking, Superannuation and Financial Services Industry (Hayne Inquiry) for improvements in ASIC’s litigation approach and enforcement policy.
The thesis includes an empirical study of ASIC’s enforcement priorities and praxis through an analysis of new enforcement activities within a defined period, 1 January 2019 to 31 December 2023. This study reveals what ASIC did on the ground, how closely ASIC hewed to the Hayne Inquiry’s recommendations and exposes the weaknesses in ASIC’s subsequent enforcement priorities and praxis. The thesis argues that ASIC’s response to the Hayne Inquiry’s recommendations for improvements in ASIC’s litigation approach and enforcement policy was appropriate in that it produced an evolution in ASIC’s enforcement priorities and praxis. Such an evolution, however, came about through ASIC’s adoption of a forcefully litigious approach and a high intensity of enforcement activities in the period between 2019 and 2023, within one industry in the Australian economy in particular, being the financial services industry. The thesis argues that it is important to ensure that in the future ASIC should not be preoccupied with one industry to the detriment of its enforcement of the economy as a whole
Synthesis and Investigation of Multi-Phase DNA Microspheres
No full textThis thesis presents a new method for the formation of DNA nanostar (NS) synthetic cells, with chemically and structurally distinct core and shell regions. First, modifications to the physical structure of nanostars were investigated. The inter-related effects of NS size, shape, inter-star binding strength, and environmental conditions on coacervate properties were investigated. A method for the experimental determination of NS phase-separation temperature was presented.
Orthogonal nanostars were combined in solution and shown to form distinct populations of coacervates. Selective modification of one NS population to induce limited interaction with the other resulted in increasing interfacial contact. The degree of contact was dependent on the energy profiles of each nanostar and the strength of inter-phase complementarity. Techniques for the prediction and formation of patchy, two-layer, and three-layer coacervates were presented, with each phase consisting of distinct nanostars. The majority of experimental micro-architectures aligned with predictions. Effects of key NS variables on the resulting coacervate micro-architecture were investigated.
Nanostar design allowed for control over the physical properties of the micro-scale coacervates. The encapsulation of a nanostar droplet in a two-layer core-shell system was shown to prevent core fusion over time at elevated temperatures. The targeted capture of cargo in multi-phase coacervates was demonstrated, using ssDNA-functionalised biomolecules. The location and degree of capture was controlled through the selective functionalisation of the cargo, along with selective release of cargo through toehold-mediated strand displacement reactions. The shell of a core-shell coacervate was shown to selectively prevent the infiltration of core-targeted cargo, acting as a switchable filter. Finally, proof-of concept designs were presented for the dynamic switching of cargo compartmentalisation and coacervate micro-architecture
A Comprehensive Exploration of Video Understanding: Perspectives on Sampling, Backbone, Representation, and Cross-Modal Learning
No full textVideo understanding is a fundamental area in computer vision with applications in autonomous driving, security, healthcare, and entertainment. As video becomes a dominant medium for information exchange, automatic interpretation is increasingly essential. While action recognition has been the foundation of video understanding, recent multimodal advancements have expanded the field to tasks like video-text matching and video question answering.
This thesis explores key advancements across multiple dimensions. First, it optimizes video recognition through salient frame selection, introducing Non-saliency Suppression Network (NSNet) to enhance efficiency and accuracy. It then investigates video model backbones, proposing the Arithmetic Temporal Module (ATM)—a plug-and-play component for temporal modeling compatible with CNNs and vision transformers.
In self-supervised video representation learning, the Macro-to-Micro Semantic Correspondence (MaMiCo) task improves learning in the absence of labeled data. Moving toward weakly supervised learning, the Text4Vis framework adapts vision-language models for video recognition by leveraging text embeddings as classifiers, enhancing zero-/few-shot recognition.
To incorporate real-world textual metadata, Cap4Video demonstrates that auxiliary captions improve text-video retrieval and recognition. Lastly, addressing the gap in video-based multimodal large language models (MLLMs), this thesis introduces Dense Connector, a plug-and-play module that enhances vision-language integration, and FreeVA, a training-free extension of image-based MLLMs to video, achieving state-of-the-art performance.
These contributions collectively advance video understanding, providing insights and practical approaches for future research and applications
Building Good Jobs for Women on the Central Coast
No full textWomen on the Central Coast want good jobs that are financially rewarding, provide opportunities for training and progression, and allow them to meet their family commitments, and health and wellbeing needs across all stages of life. However, living on the Central Coast can limit women’s access to jobs that match their skills and often imposes a high time penalty on women who commute to Sydney or Newcastle for good jobs. On the Central Coast and its environs, poor public transport, inaccessible and high-cost early childhood education and care, combined with a lack of good flexible work and training opportunities leave many women with limited opportunities and constrained choices around what work they can pursue alongside family and personal care needs.
This report analyses how women on the Central Coast experience work now, and what they aspire to in their working futures. Barriers and enablers of work vary by stage of life and career, as well as by family type. With limited options, women do their best to balance work and care, but this often comes at significant cost to their own wellbeing as they manage the ‘mental load’ of balancing these competing tasks. This is unfair and inefficient. It limits women’s opportunities for economic security and puts a brake on economic productivity in the region.
Public investment on the Central Coast in essential public services such as public transport and early childhood education and care, alongside expanded opportunities for high quality flexible and part-time work, will change and improve employment prospects for women and support their aspirations for a successful future at work. Improving opportunities for women’s employment will benefit households, business and government. It will provide an important
boost for the Central Coast economy, promote productivity and improve women’s economic security
UOS2301-003RTX Overseas wheat phenotyping 2024/25
No full textA total of 519 lines were tested in Kenya against stem rust (Ug99 lineage pathotypes – TTKSK, TTKST, TTKTT, TTTTF and TTKTT+Sr8115B) and stripe rust (race PstS16). The tested lines were derived from two resistant genotypes: Kenya Kudu (an east African line) and Kingbird (a CIMMYT line released in Ethiopia) both of which have showed good levels of resistance (characterized and/or uncharacterized) to all three rusts in field trials at PBI Cobbitty. These lines were crossed with Avocet to develop DH populations. High quality data was generated on these lines with two readings each for stem rust and stripe rust at 10-day interval (data file attached).Wheat stem rust (WSR) and stripe rust (WYR) screening was conducted at Nojoro, Kenya, the field plots comprised 1 m paired rows sown on top of 0.8 m wide raised beds. For stem rust screening at Kenya, a mixture of urediniospores of races in the Ug99 lineage (TTKSK, TTKST, TTKTT, TTTTF and TTKTT+Sr8115B) were inoculated onto susceptible spreaders (Duma and Chozi) to facilitate uniform disease development. Although the primary purpose of screening in Kenya is stem rust, data was also collected for stripe rust that appeared naturally in field (primarily race Pst16).
Disease severity was recorded according to a modified Cobb Scale by recording disease score [0-100%] and host response [R = Resistant; MR = moderately resistant; MR-MS = moderately resistant to moderately susceptible; MS = moderately susceptible; S = susceptible] when the severity on the respective susceptible check varieties reached 90-100S
Zieria obcordata DArT data
No full textData supporting the analysis of Zieria obcordata in the paper "Conservation genomics uncovers disjunct subspecies and critically low diversity in Zieria obcordata A.Cunn. (Rutaceae)